Imposing constraints on weights of input layer in NN

Question

I want to train a NN but I want the weights of the input layer to obey many constraints, more specifically:

• individual weights should be less than a given value